Hope For The Muddled Masses

It's enough to make online retailers shudder. About half of all people who visit a commercial Web site intending to buy something give up because, above all, they are confused—by product descriptions, navigation and checkout procedures. To help them out, some online retailers offer shoppers "chats"—real-time typed dialogues via the Internet—with sales agents, but these haven't worked very well. Most visitors don't bother clicking chat-invitation buttons, and many who do leave after a frustrating wait. Besides, chats are expensive—they cost a retailer about $5 a pop—and only a small minority of visitors are serious shoppers. "Random isn't profitable," says Laurent Deslandres, senior Web expert at the Paris consultancy Nexstage.

Now companies are learning to profile visitors on the sly, targeting those who are confused (and therefore most likely to leave the site). These include people who shuttle repeatedly between Web pages that describe similar products, scroll up and down several times, linger or, even worse, open a new window to reread the description of a product in the shopping cart. These tip-offs identify shoppers "obviously having trouble making a decision—that's when we would intervene," says Ravi Vijayraghavan, head software engineer at 24/7 Customer, a large software developer and contact center in Bangalore, India, that plans to offer a "confusion index" service for online retailers later this year.

The largest company operating confusion-indexing programs, sometimes called proactive chatting, is New York-based LivePerson. Its 7,000 clients include Apple, AT&T, Bank of America, British Telecom, the Royal Bank of Scotland and Verizon. LivePerson, like 24/7 Customer, operates the software free of charge, taking a cut of sales it helps close. Last year the company took in $54 million in revenues; this year revenue is on track to top $75 million.

The more factors analyzed, the more effective the service. A confusion ranking rises if a shopper fills in forms slowly. It rises faster if a shopper for, say, a laptop hesitates before typing answers to questions that require a great deal of comprehension (what bus speed? how much RAM?). The confusion ranking can skyrocket if the shopper is in Montana or Wyoming, the two least computer-savvy U.S. states, according to 24/7 Customer. The scale runs 1 to 100—lucid to bewildered.

Circumstances provide clues even before a customer starts navigating. Shoppers in Silicon Valley are apparently well versed in contemporary interior design. They start off with a low confusion ranking when landing on sites offering minimalist furniture. Shoppers in America's Midwest are less confused than those from coastal states when browsing for jewelry and watches. Confusion levels rise considerably, across the board, before Christmas and Mother's and Father's days, when gift givers venture into unfamiliar product categories. If a shopper is using a slow dial-up Internet connection, that tends to indicate limited online experience, which nudges up the confusion ranking—and therefore the money-making potential of offering chat assistance.

If chat-assisted shopping takes off, agents will likely be in short supply. The industry turns over quickly. Of the 600,000 people who work phones in U.K. call centers, a quarter quit every year. "We're running out of people in the U.K. who will do this," says Steve Morrell, director of ContactBabel, a contact-center consultancy in Sedgefield, England. Attrition levels in developing countries like India are high, too.

One way to get around the labor shortage is to offer automated chats. Retailers have had some success with software that answers questions based on keywords. Artificial Solutions, a Barcelona designer and operator of automated chats, is developing a system that will start out by assuming that visitors who stumble in to a site from a search-engine results page are more likely to be disoriented than someone who has typed in the Internet address, who is likely to be familiar with the site. The automated dialogue that follows will take such factors into account. The biggest virtue of an automated chat, of course, is that it is inexpensive—each one costs about $1. Artificial Solutions may be on to something: this year the firm expects to double its $9.84 million 2007 revenues.

Many online merchants are reluctant to trust automated chats with the formidable task of reducing shopping-cart abandonment—the bane of the industry. As the checkout process moves forward, confusion often morphs into nervousness, especially when goods are expensive. Here the ability to match skilled and psychology-savvy chat agents with the most confused shoppers is especially profitable. "Only a human can sense your nervousness," says Peter Samuelsen, CEO of Novomind, a Hamburg, Germany, developer of both automated and nonautomated chat software. Humans can provide the "reassurance needed so that your nervous finger will still press 'submit'." Adept Web surfers may dislike the idea of being monitored for confusion; for the muddle-minded, being belittled with a high confusion score may be a godsend.